A Low Complexity Multivariate Regression Based Flood Forecasting Model Using an Optimized WSN Deployment Scheme
Arnab Raha, Shovan Maity, Souvik Kr. Mitra, Ramesh Jana, Mrinal Kanti Naskar · Advanced materials research · 2011
Very recently work has been done to develop efficient disaster forecasting systems utilizing WSN technology. Such networks pose a tremendous design challenge such as the ability to cope with node failure, limited power, distributed prediction, wide variety of sensors and the need for communication over a large area. Our paper introduces a Predictive Environmental Sensor Network (PESN) Architecture which employs a minimal deployment scheme to ensure connectivity among the nodes involved within the network. On this connected network we run our distributed statistical model for forecasting. The statistical process used for this real time prediction uses multiple variable regression method providing the advantages of simplicity and robustness much needed in low power and limited ability sensor nodes. The versatility of the forecasting model is proved on its independence on the number of parameters, as it can incorporate as many variables into the algorithm as required, as long as there is sufficient positive correlation with the instantaneous river water level.